Abstract
Cancer cell–derived extracellular vesicles (EVs) have unique protein profiles, making them promising targets as disease biomarkers. High-grade serous ovarian carcinoma (HGSOC) is the deadly subtype of epithelial ovarian cancer, and we aimed to identify HGSOC-specific membrane proteins. Small EVs (sEVs) and medium/large EVs (m/lEVs) from cell lines or patient serum and ascites were analyzed by LC-MS/MS, revealing that both EV subtypes had unique proteomic characteristics. Multivalidation steps identified FRα, Claudin-3, and TACSTD2 as HGSOC-specific sEV proteins, but m/lEV-associated candidates were not identified. In addition, for using a simple-to-use microfluidic device for EV isolation, polyketone-coated nanowires (pNWs) were developed, which efficiently purify sEVs from biofluids. Multiplexed array assays of sEVs isolated by pNW showed specific detectability in cancer patients and predicted clinical status. In summary, the HGSOC-specific marker detection by pNW are a promising platform as clinical biomarkers, and these insights provide detailed proteomic aspects of diverse EVs in HGSOC patients.
INTRODUCTION
Extracellular vesicles (EVs), long known to circulate in human body fluids, were recently recognized as essential tools for intercellular communications and promising disease biomarkers (1, 2). Among the various bioactive cargoes in EVs, membrane proteins are of the highest importance as targets for elucidating subtypes of EVs and for considering further applications (3). Although EVs in body fluids are quite heterogeneous, highly specific EV membrane protein markers enable capturing specific EVs (4). With advanced technologies, such as imaging flow cytometry or antibody-based microfluidic systems, it has become possible to analyze specific EVs (5). However, as rigorous assessments for EV proteomics have been a major challenge in the field, disease-specific EV markers are still widely lacking.
Epithelial ovarian cancer (EOC) is one of the highly lethal gynecological cancers, ranked as the seventh most common cause of cancer deaths among women in the world (6, 7). Because of the difficulty of early-stage screening, most cases are diagnosed at an advanced stage, with a 5-year survival rate of less than 45% (8). EOC consists of diverse subtypes, of which high-grade serous ovarian carcinoma (HGSOC) accounts for approximately 75% of the cases, with a fatality rate of nearly 90% (9). There has been no specific and sensitive biomarker for HGSOC, and CA125, which is most commonly used in clinical diagnosis, has proven not to be effective for early detection of ovarian cancer (10). Recently, the possibility of using circulating EV–microRNA (miRNA) for EOC screening has emerged, and the concept of EV biomarkers for EOC has been proved as reasonable (11). However, the recognition of EV heterogeneity has been emerging as an essential topic only in the most recent several years, and the proteomic characterization of EOC-derived EVs has not been reported. EpCAM and CD24 have often been used as ovarian cancer EV markers, despite accumulating prior works that they are not specific for EOC (12). HGSOC-specific EV protein markers with practical utility have not been identified.
Clinical applications require EV isolation methods that are simple, and nanotechnology-based device innovations have been discussed (13). One approach is to use microfluidic systems to design devices with the advantage of making EV preparation easier (14). Our group has been engaged in the development of microfluidic devices for collecting EV-encapsulated miRNAs from urine (15, 16), using zinc oxide (ZnO) nanowires (NWs) that were prepared by hydrothermal synthesis (17). While this device has a promising application for isolating EV from urine (18), it was not successful in serum and ascites where the ZnO NWs were shown to be dissolved by foreign ions (19), partly due to the presence of highly acidic and coordinately unsaturated zinc sites. For this reason, further optimization is required to purify intact EVs from serum or ascites, and molecular modification of the surface of ZnO NWs was considered to resolve these issues.
Here, we present three notable findings supporting EV-based EOC biomarkers for clinical uses. First, we investigated the size distribution of EVs and found that small EVs (sEVs) were better targets as biomarkers than medium/large EVs (m/lEVs). We then performed shotgun proteomics by using liquid chromatography–tandem mass spectrometry (LC-MS/MS) to reveal unique protein profiles for each EV size class with substantial functional differences between small and large EVs. Second, we found HGSOC-associated sEV protein markers with features desirable for clinical utility. The LC-MS/MS for the sample set identified HGSOC-related proteins, of which three markers were selected and validated. Third, we innovated a novel sEV isolation method, named polyketone-coated NWs (pNWs), and showed that it can be used to isolate sEVs from serum or ascites in a simple procedure. Together, our results are evidence that the newly identified HGSOC-associated protein markers are promising targets both as biomarkers themselves and in applications for isolating specific EVs in circulating body fluids.
RESULTS
Isolation and characterization of EVs
To obtain globally applicable HGSOC-associated EV proteomic data, sEVs and m/lEVs were concurrently isolated from 10 cell lines, including HGSOC-like ovarian cancer cells, OVCAR3, KURAMOCHI, OV90, CAOV3, COV362, and SKOV3, (20) and noncancer cell lines, immortalized human ovarian surface epithelium cell line HOSE1/2, immortalized fibroblast cell line HFF2, and immortalized mesothelial cell line MTK, as described in Materials and Methods (Fig. 1A). EVs were also isolated from sera and ascites of HGSOC patients. To control for EV quality, nanoparticle tracking assays (NTAs) were performed, determining the precise size of EVs and confirming the presence of sEVs (with a size of around 100 nm diameter) and m/lEVs (those with larger size) in cell culture samples (Fig. 1B and fig. S1A). NTAs were also performed in clinical samples, and the results were similar to cell culture supernatant EVs (Fig. 1C). We performed cryo–electron microscopic analyses of KURAMOCHI cells to visualize EVs and successfully observed the vesicles with lipid bilayer (Fig. 1D). To further characterize EVs, immunoblotting analyses for the conventional EV makers, CD9/CD63/CD81, and the non-EV marker, GRP, were performed using the same amount of EVs in each sample (Fig. 1, E and F). The CD63-positive smear bands indicating CD63 in sEVs were consistent with prior reports. GRP positivity was observed for total cell lysate (TCL) and some m/lEVs. Regarding CD9/81, diverse intensities of bands were observed in EVs from cell lines and clinical samples. This may be affected by the condition of resources because body fluids contain various EVs from diverse origins. Before shotgun proteomics, we removed albumin and immunoglobulin G (IgG) from the EV samples derived from serum and ascites because those abundant proteins can mask the less-abundant proteins of interest in our samples (fig. S1B). After performing global MS analyses, first we checked for the presence of the EV-associated proteins of categories 1 to 5 of the MISEV2018 indications (Fig. 1, G and H) (21). The International Society for Extracellular Vesicles (ISEV) highlighted these categories as indicative of the quality of purified EVs, with the following constituents: transmembrane or glycosylphosphatidylinositol-anchored proteins (category 1), cytosolic proteins (category 2), major constituents of non-EV structures (category 3), larger EV-associated proteins (category 4), and secreted or luminal proteins (category 5). The positivity for the proteins of category 1 was relatively greater in sEV than m/lEV samples, and there was no obvious difference in cancer-derived or non–cancer-derived EVs. In summary, we precisely isolated sEVs and m/lEVs from cell culture supernatants and HGSOC patient samples, and obtained their proteomic profiles.
Profiling sEVs and m/lEVs
After MS analysis confirmed the presence of EV-associated proteins in our EV samples, we proceeded to check for the presence of EV proteins previously reported associated with EOC and especially with HGSOC. For the set of candidate proteins, we selected 59 EV proteins; 42 proteins were reportedly detectable in EOC cell line samples (14, 22–24), and 26 were present in EV samples of EOC patients (Fig. 2, A and B). We also selected, for a similar analysis, 165 proteins associated with ovarian cancer according to the UniProt Protein database (http://uniprot.org/) (fig. S2). The most frequently reported marker for EOC, EpCAM, does not have sufficient sensitivity and specificity for ovarian cancer in clinical practice, and it is not specific to ovarian cancer (25). Other well-known EOC-associated proteins, such as CA125 or mesothelin, are specifically highly present in cancer samples but do not depend on EVs. Thus, although the proteomic analyses worked well, markers more specific for HGSOC than those in earlier reports were needed to be identified….
